FDI and environmental regulations in China

Abstract

This paper uses provincial socioeconomic and environmental data and investigates whether there exists an intra-county pollution haven effect for China. We examine whether differences in the stringency of environmental regulations affect the choice of location for FDI in China. We use a five-year panel dataset for 30 provinces in China that includes three measures of environmental regulations, which vary across time and province, and a significant number of control variables including measures of agglomeration and factor abundance. We control for unobserved heterogeneity by using a feasible generalised least square estimator. Our results suggest that environmental stringency has a significant and negative effect on FDI, leading us to conclude that, ceteris paribus, FDI prefers to locate into regions with relatively weak environmental regulations. This provides some support for the existence of a pollution haven within China.

Keywords:

• FDI
• environmental regulations
• pollution haven hypothesis

JEL Classifications:

• F21
• O13
• Q58

Acknowledgment

The research of this paper is funded by the Leverhulme Trust grant number F/00094/AG. The authors gratefully acknowledge the support of Dr Robert J.R. Elliott and Dr Matthew A. Cole from University of Birmingham, UK. They also acknowledge the helpful comments from anonymous referees.

Notes

∗,

∗∗, and

∗∗∗ indicate significant at 10%, 5% and 1% level, respectively.

∗,

∗∗, and

∗∗∗ indicate significant at 10%, 5% and 1% level, respectively.

∗,

∗∗ and

∗∗∗ indicate significant at 10%, 5% and 1% level, respectively.

1. Chinese industry is split into the following three main categories. Primary industry refers to extraction of natural resources, i.e. agriculture (including farming, forestry, animal husbandry and fishery). Secondary industry involves processing of primary products, i.e. industry (including mining and quarrying, manufacturing, production and supply of electricity power, gas and water) and construction. Tertiary industry refers to all other economic activities not included in the primary and secondary industries.

2. The enterprises are all state-owned and non-state-owned enterprises above the designated size, which refers to enterprises with an annual sales income of over 5 million RMB yuan (about US$0.60 million).

3. After taking logs, the skewnesses of all the variables are significantly reduced. Results for the levels models are available from the authors on request.

4. The table for descriptive statistics is available from the authors on request.

5. The between estimator is obtained by using OLS to estimate the models which use the time-averages for both dependent and independent variables and then run a cross-sectional regression (Wooldridge 2000, Chapter 14, p. 442). GLS estimators produce more efficient results than between estimators because they use both the within and between information.

6. Available from the authors on request.

7. We also estimate the models using an alternative method, i.e. OLS models with panel-corrected standard errors (PCSEs), to control for both heteroscedasticity and autocorrelation. The results are broadly similar to those using FGLS. Results are available from the authors on request.

8. The results of regressions with random effects are available from the authors on request.

9. The major syntaxes include xtreg with fe and re, xtgls, and xtpcse.

10. The results for sensitivity analysis and levels estimations are available from the authors on request.

11. The turning point is around 3000–4000 RMB, which is lower than the current income level in China, indicating that labour costs deter FDI inflows at the current income level in China.

12. We also estimate our regressions using the numbers of enterprises as a proxy of agglomeration. These enterprises include all state-owned and non-state-owned industrial enterprises with an annual sales income of over 5 million RMB yuan. Our main results are unaffected.

13. We also estimate our regressions including railway density and road density separately but the results were very similar. We also estimated the regressions respectively, including the numbers of ports in each province and the dummy variable for coastal provinces. Both coefficients are positive but not significant.